Action Budgets, Iteration Caps, and Cost Governors

A mid-sized e-commerce agent's monthly LLM cost jumped from $1,200 to $4,800 after its team enabled the "order-tracking" skill. That is not a pricing bug. That is an agent that found a new loop and kept spending inside it. Microsoft's Agent Governance Toolkit (April 2, 2026) codifies the defense against this class: per-request `max_tokens`, per-task token and dollar budgets, per-day/month caps, iteration caps, tiered model routing, prompt caching, context windowing, HITL checkpoints on expensive actions, kill switches on budget breach. Anthropic's Claude Code Agent SDK ships the same primitives under different names. Financial velocity limits — e.g. cut access on >$50 in 10 minutes — catch loops faster than monthly caps.

Learn 1 hr Python (stdlib, layered cost-governor simulator)

Prerequisites: Phase 15 · 10 (Permission modes), Phase 15 · 12 (Durable execution)

#The Problem

Autonomous agents spend real money on every turn. A chatbot's bad output is a bad reply; an agent's bad loop is a bill. The industry-documented term for the failure mode is "Denial of Wallet" — the agent keeps reasoning, keeps tool-calling, keeps billing, and nothing stops it because nothing was designed to.

The fix is not one number. It is a stack of limits at different time scales and granularities: per-request, per-task, per-hour, per-day, per-month. A well-designed stack catches a runaway loop within minutes, a slow leak within hours, and a bad release within a day. The same stack keeps a budget at all when the agent is long-horizon and autonomous.

This is an engineering lesson: the math is trivial, the discipline is where teams fail. The list of limits below is all named either in the Microsoft Agent Governance Toolkit or the Anthropic Claude Code Agent SDK docs.

#The Problem

Keep going. The rest of this lesson is one step away.